CN109712168A - A kind of method for tracking target and electronic equipment, device and storage medium - Google Patents
A kind of method for tracking target and electronic equipment, device and storage medium Download PDFInfo
- Publication number
- CN109712168A CN109712168A CN201811355222.XA CN201811355222A CN109712168A CN 109712168 A CN109712168 A CN 109712168A CN 201811355222 A CN201811355222 A CN 201811355222A CN 109712168 A CN109712168 A CN 109712168A
- Authority
- CN
- China
- Prior art keywords
- target
- tracked
- characteristic point
- blocked
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The embodiment of the present application provides a kind of method for tracking target and electronic equipment, device and storage medium, which comprises determines the characteristic point at the position of target to be tracked not being blocked;According to the characteristic point, the occlusion state of the target to be tracked is determined;According to the occlusion state of the target to be tracked, the corresponding feature templates of the target to be tracked are adjusted;In the corresponding target area of the target to be tracked, searches for and track the region to match with feature templates adjusted;Wherein, the target area is for tracking the mesh target area to be tracked.
Description
Technical field
This application involves field of image detection more particularly to a kind of method for tracking target and electronic equipment, device and storage
Medium.
Background technique
Target occlusion to be tracked is all a difficulties in target tracking algorism all the time, because in practical application
In, occlusion issue can often occur in object tracking process, if algorithm cannot handle occlusion issue, the practicality well
It will have a greatly reduced quality, therefore the detection blocked is most important for track algorithm.
Although in the related technology track algorithm can long-time stable track target to be tracked, its premise set up
Be: target to be tracked is not blocked.If target to be tracked is blocked, while track algorithm does not know that target to be tracked is hidden
Gear, then algorithm will continue to continue according to original mechanism, at this moment track algorithm can background as prospect carry out with
Track causes target to be tracked with losing so that the tracing positional of target to be tracked deviates.
Summary of the invention
In view of this, the embodiment of the present invention to solve problems of the prior art, provides a kind of method for tracking target
And electronic equipment, device and storage medium.
The technical solution of the embodiment of the present application is achieved in that
The embodiment of the present application provides a kind of method for tracking target, which comprises determines not hidden for target to be tracked
The characteristic point at the position of gear;According to the characteristic point, the occlusion state of the target to be tracked is determined;According to the mesh to be tracked
Target occlusion state adjusts the corresponding feature templates of the target to be tracked;In the corresponding target area of the target to be tracked
In, it searches for and tracks the region to match with feature templates adjusted;Wherein, the target area be for track it is described to
Track mesh target area.
The embodiment of the present application provides a kind of target tracker, and described device includes: the first determining module, the second determining mould
Block, the first adjustment module and the first search module, in which: first determining module, for determine target to be tracked not by
The characteristic point at the position blocked;Second determining module, for determining the screening of the target to be tracked according to the characteristic point
Gear state;The first adjustment module adjusts the target pair to be tracked for the occlusion state according to the target to be tracked
The feature templates answered;First search module, for searching for and tracking in the corresponding target area of the target to be tracked
The region to match with feature templates adjusted;Wherein, the target area is the area for tracking the target to be tracked
Domain.
The embodiment of the present application provides a kind of electronic equipment, including memory and processor, and the memory is stored with can be
The computer program run on processor, the processor realize the step in above-mentioned method for tracking target when executing described program
Suddenly.
Accordingly, the embodiment of the present application provides a kind of computer readable storage medium, is stored thereon with computer program, should
The step in above-mentioned method for tracking target is realized when computer program is executed by processor.
The embodiment of the present application provides a kind of method for tracking target and electronic equipment, device and storage medium, wherein firstly,
Determine the characteristic point at the position of target to be tracked not being blocked;According to the characteristic point, the screening of the target to be tracked is determined
Gear state;Then, according to the occlusion state of the target to be tracked, the corresponding feature templates of the target to be tracked are adjusted;Most
Afterwards, it in the corresponding target area of the target to be tracked, searches for and tracks the region to match with feature templates adjusted;
In this way, according to the circumstance of occlusion of target to be tracked, adjust the different templates for tracking, so as to flexibly to pedestrian into
Line trace especially has very strong robustness to partial occlusion, greatly in this way, obtaining the target tracking algorism of long-time stable
Big reduce occurs the case where target to be tracked is with losing because of partial occlusion.
Detailed description of the invention
In attached drawing (it is not necessarily drawn to scale), similar appended drawing reference can describe phase in different views
As component.Similar reference numerals with different letter suffix can indicate the different examples of similar component.Attached drawing with example and
Unrestricted mode generally shows each embodiment discussed herein.
Figure 1A is the flow diagram of the embodiment of the present application method for tracking target;
Figure 1B is the flow diagram of the another method for tracking target of the embodiment of the present application;
Fig. 1 C is the flow diagram of another method for tracking target of the embodiment of the present application;
Fig. 1 D is the flow diagram of the embodiment of the present application still another object tracking;
Fig. 1 E is the flow diagram of another method for tracking target of the embodiment of the present application;
Fig. 1 F is the flow diagram of another method for tracking target of the embodiment of the present application;
Fig. 2 is the another flow diagram for realizing method for tracking target of the embodiment of the present application;
Fig. 3 is the application scenarios schematic diagram of the embodiment of the present application method for tracking target;
Fig. 4 is the composed structure schematic diagram of the embodiment of the present application target tracker;
Fig. 5 is a kind of hardware entities schematic diagram of electronic equipment in the embodiment of the present application.
Specific embodiment
It should be appreciated that specific embodiment described herein is only used to explain the application, it is not used to limit the application.
In subsequent description, it is only using the suffix for indicating such as " module ", " component " or " unit " of element
The explanation for being conducive to the application, itself does not have a specific meaning.Therefore, " module ", " component " or " unit " can mix
Ground uses.
Electronic equipment can be implemented in a variety of manners.For example, electronic equipment described in this application may include such as
The mobile electrons such as personal digital assistant (Personal Digital Assistant, PDA), navigation device, wearable device are set
It is standby, and the stationary electronic devices such as number TV, desktop computer.
It will be illustrated by taking mobile electronic device as an example in subsequent descriptions, it will be appreciated by those skilled in the art that in addition to
It is used in particular for except the element of mobile purpose, can also apply to the electricity of fixed type according to the construction of presently filed embodiment
Sub- equipment.
In the related art, track algorithm is largely all blocked using pedestrian detection algorithm to detect, when detecting pedestrian
After continue to track, do not detect, tracking terminate.This mode can not solve blocking for target arbitrary portion to be tracked
Problem, when such as blocking 2/3 or 4/5;Current pedestrian detection algorithm can only realize preassigned trained fixed quantity
Classifier, such as three detectors of training: upper body detector, lower part of the body detector, whole body detector;Although these can be preferable
It handles these three to block, but is blocked if there is the 2/3 of target to be tracked or 1/2 blocks that (but shield portions are 1/4
To 3/4 part) when, then can not be detected using these three types of detectors, at this time track algorithm will be considered that target to be tracked with
It loses, stops tracking, to cannot achieve long-time stable tracking.
To solve in the case that target part to be tracked is blocked in the related technology, can not long-time stable tracking ask
Topic, the embodiment of the present application provide a kind of method for tracking target, and Figure 1A is that the process of the embodiment of the present application method for tracking target is illustrated
Figure, as shown in Figure 1A, the described method comprises the following steps:
Step S101 determines the characteristic point at the position of target to be tracked not being blocked.
Here, target to be tracked can be any pedestrian for needing to track, the position not being blocked, it is possible to understand that
For, during tracking target to be tracked, the position that the object in actual scene during not being tracked is blocked, than
Such as, the certain positions that do not blocked by tree, billboard or other pedestrians etc..The position not being blocked, it is also understood that
For the position for the target to be tracked that can be identified to;The characteristic point is that the position of target to be tracked not being blocked is certain
Point, for example, the head of target to be tracked, shoulder, waist, leg, foot etc. take 18 characteristic points altogether, according to the spy not being blocked
Point is levied, that is, can determine the position of target to be tracked not being blocked.Characteristic point is the characteristic point on head, shows target to be tracked
The position not being blocked is head;The method is applied in electronic equipment, which includes acquisition device, the acquisition
Device can be the camera of electronic equipment, the characteristic point at the position not being blocked for acquiring target to be tracked.
Step S102 determines the occlusion state of the target to be tracked according to the characteristic point.
Here, the occlusion state of target to be tracked includes at least: unobstructed state, partial occlusion state and blocking shape completely
State.
Step S103 adjusts the corresponding character modules of the target to be tracked according to the occlusion state of the target to be tracked
Plate.
Here, if the occlusion state of target to be tracked is unobstructed state, fisrt feature template is created, is labeled as template
A;The fisrt feature template is created according to target feature point to be tracked, it can be understood as is comprising target to be tracked
The three-dimensional (3 D) manikin or plane manikin of whole characteristic points;So when the occlusion state of target to be tracked is partial occlusion
When, second feature template is created according to the characteristic point not being blocked;Fisrt feature template is adjusted to second feature template, second
The characteristic point that feature templates include is less than the characteristic point that the fisrt feature template includes;The second feature template can pass through
Following methods obtain:
Firstly, obtaining the boundary rectangle of characteristic point, second feature template is established using this rectangular area, described second
Feature templates, it can be understood as be the three-dimensional (3 D) manikin or plane of the characteristic point not being blocked comprising target to be tracked;
Then, switch the feature templates of track algorithm, i.e., be switched to second feature template from original fisrt feature template, continue to track
With update target position to be tracked.
Step S104 is searched for and is tracked and character modules adjusted in the corresponding target area of the target to be tracked
The region that plate matches.
Here, the target area is for tracking region shared by the body of the target to be tracked.The step
S104 in the corresponding target area of target to be tracked, is searched for it is to be understood that if target to be tracked is in unobstructed state
And track the first area that the characteristic point that the position not being blocked includes with fisrt feature template matches.The target to be tracked
Corresponding target area can be understood as the body region of target to be tracked, for example, target to be tracked is personage A, then personage A
Corresponding target area can be region corresponding to the body of personage A.Search for and feature adjusted in the target area
The most like part of template, if similarity is greater than certain threshold value, then it is assumed that target following to be tracked is normal.
In the embodiment of the present application, by the circumstance of occlusion according to tracking target, the different features for tracking is adjusted
Template, so as to flexibly be tracked to pedestrian, in this way, the tracking of long-time stable is realized, especially to tracking mesh
The case where target position is at least partially obscured has very strong robustness, greatly reduces and occurs target to be tracked because of partial occlusion
The case where with losing.
The embodiment of the present application provides a kind of method for tracking target, and Figure 1B is the another method for tracking target of the embodiment of the present application
Flow diagram the described method comprises the following steps as shown in Figure 1B:
Step S111 determines target to be tracked from image to be detected of input.
Here, described image to be detected, it can be understood as be the image comprising target to be tracked, removed in image to be detected
Except target to be tracked, it is also possible to include other pedestrians, animal, building and plant etc.;Target to be tracked, which can be, appoints
What object that can move, for example, pedestrian, animal even advance in vehicle etc.;Therefore, the mapping to be checked from input
Target to be tracked is determined as in, can be the one or more people for determining from multiple pedestrians of image to be detected and needing to track;
Target to be tracked is determined in described image to be detected from input, be can also be and is determined that animal therein is from image to be detected
Target to be tracked.For example, including five people in image to be detected, wherein one is suspect, in order to chase as soon as possible
The suspect tracks the suspect in monitor video, to determine the action of the suspect;If
Target to be tracked is animal, for example, one has just been finished the golden monkey of operation, needs to pay close attention to the body feelings of the golden monkey constantly
Condition, then being tracked by monitor to the panda.
Step S112 detects the position of the target to be tracked not being blocked, obtains the position not being blocked
Characteristic point.
Here, the step S112 determines these positions not being blocked it is to be understood that according to the position not being blocked
The characteristic point for including, for example, if target to be tracked is pedestrian, and the head of pedestrian is not blocked, determines that head is corresponding
Two characteristic points are the characteristic points at the position not being blocked;If target to be tracked is the vehicle in advancing, and entire vehicle body
It is not all blocked, then it is determined that the trade mark of the license plate number of vehicle, vehicle, vehicle etc. are the characteristic point at the position not being blocked;
Then, fisrt feature template is created according to the trade mark of vehicle, the trade mark of vehicle, vehicle etc., in order to track to vehicle;Such as
The license plate number of fruit vehicle is blocked (i.e. vehicle is at least partially obscured), according to the creation such as the trade mark for the vehicle not being blocked and vehicle the
Then two feature templates track the vehicle in traveling according to second feature template.
Above-mentioned step S111 and step S112 gives a kind of realize and " determines the position of target to be tracked not being blocked
Characteristic point " mode, in this approach, firstly, determine target to be tracked, then by detecting the position not being blocked,
So that it is determined that the corresponding characteristic point in the position, so, it is possible accurately to get the characteristic point not being blocked for tracking, into
And in tracing process, tracking will not be caused to fail because of partial occlusion.
Step S113 determines the occlusion state of target to be tracked according to the characteristic point.
Step S114 adjusts the corresponding character modules of the target to be tracked according to the occlusion state of the target to be tracked
Plate.
Step S115 is searched for and is tracked and character modules adjusted in the corresponding target area of the target to be tracked
The region that plate matches.
In the present embodiment, when the pedestrian (for example, suspect) tracked to needs is tracked, pedestrian movement's
Process can inevitably be blocked by objects such as building, trees on the way, in this case, firstly, determining that the pedestrian is not blocked
Position, be based on this, determine the characteristic point not being blocked;Then, based on the characteristic point that these are not blocked, building is comprising being somebody's turn to do
The second feature template for the characteristic point not being blocked chases after the pedestrian that position is at least partially obscured using second feature template
Track, so achievable long-time stable tracking, greatly reduces and occurs the case where target to be tracked is with losing because of partial occlusion.
The embodiment of the present application provides a kind of method for tracking target, and Fig. 1 C is another method for tracking target of the embodiment of the present application
Flow diagram according to the number of characteristic point quantity, determines the occlusion state of target to be tracked, the method as shown in Figure 1 C
The following steps are included:
Step S121 determines the characteristic point at the position of target to be tracked not being blocked.
Step S122 determines the occlusion state of the target to be tracked if characteristic point quantity is more than or equal to first threshold
For unobstructed state.
Here, if, if the quantity of characteristic point is 12, first threshold 11 is said for example, target to be tracked is pedestrian
The body region overwhelming majority of bright target to be tracked is not all blocked, so determining that the occlusion state of target to be tracked is unobstructed
State.For example, target to be tracked is sportsman, 12 features are taken altogether in the head of sportsman, shoulder, waist, leg, foot etc.
Point, only foot a characteristic point is blocked by other sportsmen, and 11 not be blocked characteristic point still can accurately show this
The identity of movement still can accurately identify the sportsman that is, according to 11 characteristic points not being blocked, so determining should
The occlusion state of sportsman is unobstructed state.
Step S123, if the characteristic point quantity is less than first threshold and is more than or equal to second threshold, determine it is described to
The occlusion state of track target is partial occlusion state.
Here, for example, first threshold 11, second threshold 5 illustrates mesh to be tracked if the quantity of characteristic point is 8
Target body region, part are not all blocked, so determining that the occlusion state of target to be tracked is partial status.For example, to
Track target is sportsman, take 12 characteristic points altogether in the head of sportsman, shoulder, waist, leg, foot etc., only foot
One characteristic point is blocked by other sportsmen, 8 not be blocked characteristic point, but according to 8 characteristic points not being blocked, still
The sportsman so can be accurately identified, so determining that the occlusion state of the sportsman is part gear state.
Step S124 determines the occlusion state of the target to be tracked if the characteristic point quantity is less than second threshold
For complete occlusion state.
Here, it if the target to be tracked is in complete occlusion state, determines that the target to be tracked is lost, terminates to chase after
Track.For example, being greater than first threshold 11 if the quantity of characteristic point is 12, determine the occlusion state of target to be tracked for all
It blocks, then, so can't detect traceable characteristic point, showing since clarification of objective point to be tracked is almost blocked completely
Therefore tracking failure when target to be tracked be in complete occlusion state, determines that target to be tracked is lost, terminates to track.For example,
Target to be tracked is cunning suspect, and during tracking, suspect is likely to conceal, such case
Under, due to that can not detect the physical feeling of suspect, then also you can't get the characteristic point for tracking, illustrate with
Track failure.
Above-mentioned step S122 to step S124, which gives, a kind of realizes " occlusion state for determining the target to be tracked "
Mode according to the quantity for the characteristic point that target to be tracked is blocked, block shape in this approach come determine target to be tracked
State so, it is possible accurately to reflect, the position that target to be tracked is blocked, determine whether to need to create second feature
Template, in order to be tracked for a long time to target to be tracked.
Step S125 adjusts the corresponding character modules of the target to be tracked according to the occlusion state of the target to be tracked
Plate.
Step S126 is searched for and is tracked and character modules adjusted in the corresponding target area of the target to be tracked
The region that plate matches.
In the present embodiment, the quantity for the characteristic point being blocked according to target to be tracked, to determine the screening of target to be tracked
Gear state creates second feature mould according to the Partial Feature point not being blocked when target to be tracked is that part is blocked
Plate, in order to continue to track, but when target to be tracked is all blocked, due to can not detect the spy not being blocked
Point is levied, then terminating to track, avoids waste tracking resource.
It in other embodiments, can be according to locating for characteristic point for the determination of the occlusion state of the target to be tracked
Position, to determine that the occlusion state of target to be tracked, process are as follows:
The position according to locating for characteristic point determines the position that is not blocked of target to be tracked, so that it is determined that target to be tracked
Occlusion state;For example, 13 characteristic points are taken altogether in the head of pedestrian to be tracked, shoulder, waist, leg, foot etc., if
Characteristic point is whole characteristic points on head and the characteristic point of shoulder, illustrates that the head of pedestrian to be tracked and shoulder are not hidden
Gear, i.e., the part of the body region of pedestrian to be tracked are blocked, accordingly, it is determined that the occlusion state of target to be tracked is that part hides
Gear.
The embodiment of the present application provides a kind of method for tracking target, and Fig. 1 D is the embodiment of the present application still another object tracking
Flow diagram, as shown in figure iD, if target to be tracked is in unobstructed state, the method can pass through following steps reality
It is existing:
Step S131 determines the characteristic point at the position of target to be tracked not being blocked.
Step S132 determines the occlusion state of the target to be tracked according to the characteristic point.
Step S133, if target to be tracked is in unobstructed state, according to the target to be tracked for being in unobstructed state
Characteristic point, create fisrt feature template.
Here, according to whole characteristic points of the target to be tracked in unobstructed state, creation includes these characteristic points
Three-dimensional or plane body templates, by the characteristic point for including in the fisrt feature template and target to be tracked during tracking
The characteristic point at the position not being blocked is compared, to determine whether that tracking is normal.For example, target to be tracked is sportsman,
Head, shoulder, waist, leg, the foot etc. of sportsman takes 12 characteristic points altogether, if the occlusion state of the sportsman is no screening
Gear state, i.e., according to this 12 characteristic points, creation includes the three-dimensional or plane body templates of these characteristic points, is then based on this
Three-dimensional or plane body templates tracks sportsman.
Step S134, in the corresponding target area of the target to be tracked, search for and track the position that is not blocked with
The first area that the characteristic point that the fisrt feature template includes matches.
Here, in the target area, search for and track the characteristic point in the position not being blocked and the fisrt feature mould
The characteristic point similarity that plate includes is greater than the characteristic point of certain threshold value, and corresponding region in the target area, is searched in simple terms
The similarity for the characteristic point which characteristic point includes with fisrt feature template in the position that Suo Wei is blocked is higher, these are similar
It spends region shared by higher characteristic point and is determined as first area;For example, in the position not being blocked shoulder characteristic point, with
The similarity for the characteristic point that one feature templates include is very high, determines that shoulder is first area.
Above-mentioned step S133 and step S134 give it is a kind of realize " if target to be tracked is in unobstructed state,
Target to be tracked is tracked " mode, in this approach, when target to be tracked is in unobstructed state, using comprising
The fisrt feature template of whole characteristic points is tracked, and ensure that the accuracy of tracking.
In the present embodiment, if target to be tracked is in unobstructed state, using all comprising target to be tracked
The fisrt feature template of feature tracks target to be tracked, to achieve the purpose that accurate tracking.
The embodiment of the present application provides a kind of method for tracking target, and Fig. 1 E is another method for tracking target of the embodiment of the present application
Flow diagram, as referring to figure 1E, if target to be tracked is in partial occlusion state, the method can pass through following steps
It realizes:
Step S141 determines the characteristic point at the position of target to be tracked not being blocked.
Step S142 determines the occlusion state of the target to be tracked according to the characteristic point.
Step S143, if the occlusion state of the target to be tracked is partial occlusion state, by fisrt feature template tune
Whole is second feature template.
Here, the characteristic point that the second feature template includes is less than the characteristic point that the fisrt feature template includes;Institute
State step S143, it can be understood as if the body region of target to be tracked is at least partially obscured, according to the position not being blocked
Characteristic point determines the boundary rectangle of these characteristic points, established using this rectangular area the three-dimensional comprising these characteristic points or
Person's areal model, as second feature template;Then, it is switched to second feature template from original fisrt feature template, continued
Track and update target position to be tracked.For example, target to be tracked is sportsman, in the head of sportsman, shoulder, waist, leg
Portion, foot etc. take 12 characteristic points altogether, if the occlusion state of the sportsman is partial occlusion state, i.e. basis is not blocked
8 characteristic points, creation includes the three-dimensional or plane body templates of these characteristic points, as second feature template, then second
Feature templates track sportsman.
Step S144, in the corresponding target area of the target to be tracked, search for and track the position that is not blocked with
The second area that the characteristic point that the second feature template includes matches.
Here, step S144 searches for the portion not being blocked it is to be understood that in the corresponding target area of target to be tracked
The similarity for the characteristic point that the characteristic point and second feature template of position include is more than or equal to the second area of preset threshold.That is, will
The characteristic point at the biggish position not being blocked of similarity for the characteristic point for including with second feature template, corresponding region, really
It is set to second area.
Above-mentioned step S143 and step S144 gives a kind of realize " if target to be tracked is in partial occlusion shape
State tracks target to be tracked " mode when target to be tracked is in partial occlusion state, will use in this approach
It is adjusted to the second feature template comprising Partial Feature point from the fisrt feature template comprising whole characteristic points in the template of tracking,
It is tracked, so according to the occlusion state of target to be tracked, feature templates of the flexible switching for tracking, to ensure that
Even if target to be tracked is at least partially obscured, also can accurately be tracked.
In the present embodiment, the occlusion state at the position based on target to be tracked, firstly, building includes the spy not being blocked
Levy the second feature template of point, then, by fisrt feature template be adjusted to second feature template in partial occlusion to
Track target is tracked, even if also can accurately be tracked so that the position of target to be tracked is at least partially obscured.
The embodiment of the present application provides a kind of method for tracking target, and Fig. 1 F is another method for tracking target of the embodiment of the present application
Flow diagram, as shown in fig. 1F, the method can be realized by following steps:
Step S151 determines the characteristic point at the position of target to be tracked not being blocked.
Step S152 determines the occlusion state of the target to be tracked according to the characteristic point.
Step S153 adjusts the corresponding character modules of the target to be tracked according to the occlusion state of the target to be tracked
Plate.
Step S154 searches for the characteristic point at the position not being blocked in the corresponding target area of the target to be tracked
The similarity for the characteristic point for including with the second feature template is more than or equal to the second area of preset threshold.
Here, during being tracked to target to be tracked, the position that is not blocked by comparing target to be tracked
Characteristic point and the second feature template characteristic point that includes similarity, to determine whether that tracking is normal, for example, target to be tracked
It is sportsman, and the characteristic point not being blocked is 8, during being tracked to the sportsman, by comparing current
It is pre- whether the similarity of the characteristic point at the position that sportsman is not blocked and the second feature template comprising this 8 characteristic points is greater than
If threshold value, if it does, illustrate that tracking is normal, otherwise, tracking failure.
Step S155, if the characteristic point that any region includes in the target area of target to be tracked, and second feature
The characteristic point that template includes mismatches, and determines that the target to be tracked is lost, terminates tracking.
Here, it if the target to be tracked is in complete occlusion state, determines that the target to be tracked is lost, terminates to chase after
Track.If the similarity for the characteristic point for not having characteristic point to include with second feature template in the target area of target to be tracked is greater than
Certain threshold value illustrates to be unable to precise search in target area to target to be tracked, i.e., BREAK TRACK to be tracked.
Above-mentioned step S154 and step S155 gives a kind of realize and " determines that the target to be tracked is lost, terminate to chase after
The mode of track ", in this approach, firstly, the feature that the characteristic point at the position not being blocked and second feature template include
The similarity of point illustrates that tracking is normal if similarity is greater than preset threshold;If similarity is less than preset threshold and illustrates to track
Abnormal, i.e., tracking is lost, and timely terminates to track;In this way, search less than traceable characteristic point when, timely terminate with
Track avoids carrying out useless tracking, wastes time and track resource.
In the present embodiment, when target to be tracked is blocked completely or the target area of target to be tracked in not with
When the higher characteristic point of feature templates similarity, determine that tracking is lost, and terminate to track in time, in order to avoid waste tracking resource.
The embodiment of the present application provides a kind of method for tracking target, and Fig. 2 is the another realization target following side of the embodiment of the present application
The flow diagram of method, as shown in Fig. 2, the described method comprises the following steps:
Step S201 determines target to be tracked from image to be detected of input.
Here, if including 5 people in image to be detected of input, target to be tracked is therefrom determined in 5 people, for example,
One of people or 2 people etc..
Step S202 detects the position of the target to be tracked not being blocked, obtains the position not being blocked
Characteristic point.
Here, if the position being blocked is head, the characteristic point on head, i.e. several characteristic points on head are obtained.
Step S203 determines blocking for the target to be tracked if the characteristic point quantity is more than or equal to first threshold
State is unobstructed state.
Here, suppose that it is provided with 13 characteristic points, first threshold 11 altogether for target whole body to be tracked, if obtain
Characteristic point is more than or equal to 11, illustrates that target to be tracked is not almost blocked, determines that the occlusion state of target to be tracked is unobstructed
State.
Step S204 creates fisrt feature template according to the clarification of objective point to be tracked for being in unobstructed state.
Here, when the occlusion state of target to be tracked is in unobstructed state, the first spy is established according to 13 characteristic points
Levy template.
Step S205, in the corresponding target area of the target to be tracked, search for and track the position that is not blocked with
The first area that the characteristic point that the fisrt feature template includes matches.
Here, the step S205 is it is to be understood that (target i.e. to be tracked includes the body of characteristic point in the target area
Region), the search part most like with the characteristic point in fisrt feature template, if similarity is greater than certain threshold value, then it is assumed that
Target following to be tracked is normal.For example, searching for spy corresponding with 16 characteristic points in fisrt feature template in the target area
The most similar part of sign point, if similarity is greater than 90%, then it is assumed that tracking is normal.
Step S206, if the characteristic point quantity is less than first threshold and is more than or equal to second threshold, determine it is described to
The occlusion state of track target is partial occlusion state.
Here, suppose that being provided with 13 characteristic points, first threshold 11 altogether for target whole body to be tracked, second threshold is
5, if characteristic point quantity is 9, greater than 5 and less than 11, illustrate that the body region part of target to be tracked is blocked, therefore, really
The occlusion state of fixed target to be tracked is partial occlusion state.
Fisrt feature template is adjusted to second feature template by step S207.
Here, the step S207 is first it is to be understood that when the occlusion state of target to be tracked is partial occlusion state
First, characteristic point is obtained, the boundary rectangle of these characteristic points is then determined, second feature mould is established using this rectangular area
Plate continues to track and update target position to be tracked finally, being switched to second feature template from original fisrt feature template.
Step S208, in the corresponding target area of the target to be tracked, search for and track the position that is not blocked with
The second area that the characteristic point that the second feature template includes matches.
Here, if characteristic point quantity is 9, the quantity for the characteristic point that second feature template includes is 9, in target area
In domain, the most like part of the characteristic point that the characteristic point and second feature template for searching for the position not being blocked include is (i.e. in mesh
It marks in region, searches for and track the secondth area that the characteristic point that the position not being blocked includes with the second feature template matches
Domain), if similarity is greater than certain threshold value (characteristic point similarity that characteristic point and second feature template in target area include
Greater than certain threshold value, illustrate in the target area comprising target to be tracked, that is, include target to be tracked), illustrate to track successfully.It is aobvious
So, if the characteristic point that any region includes in the target area of target to be tracked, the feature for including with second feature template
Point mismatches, and the characteristic point similarity for illustrating that the characteristic point for including in target area and second feature template include is lower even
0, illustrate that target to be tracked is tracked and lose, it is thus determined that target to be tracked is lost, terminates tracking.
Step S209 determines the occlusion state of the target to be tracked if the characteristic point quantity is less than second threshold
It for complete occlusion state, and determines that the target to be tracked is lost, terminates tracking.
Here, if in characteristic point including 2 characteristic points, it is less than second threshold 5, illustrates the body region of target to be tracked
It is capped exhausted big region, accordingly, it is determined that the occlusion state of target to be tracked is complete occlusion state.Due to target quilt to be tracked
It blocks completely, it is clear that tracking can not be further continued for, so determining that the target to be tracked is lost, terminate tracking.
In the embodiment of the present application, by the occlusion state according to target to be tracked, spy corresponding with occlusion state is adjusted
Levy template, so as to still be able to accurately be tracked in the case where target to be tracked is at least partially obscured, greatly reduce because
Occurs the case where target to be tracked is with losing for partial occlusion.
The embodiment of the present application provides a kind of method for tracking target, in order to which accurate detection blocks feelings to target to be tracked
Condition, the occlusion detection algorithm that the present embodiment is taken based on human body key point realize method for tracking target, comprising the following steps:
The first step determines the characteristic point of target whole body to be tracked.
Here, as shown in figure 3, at the head of target 30 to be tracked, shoulder, waist, knee and ankle, 13 are determined altogether
A characteristic point (i.e. point a to point m).
Second step, based on the characteristic point under the unobstructed state of target to be tracked, i.e., whole characteristic points establishes template A (i.e.
Fisrt feature template).
It here, include the corresponding characteristic point of 13 characteristic points in template A.
Third step distinguishes the current occlusion state of target to be tracked according to 13 human body feature points detected.
Here, the occlusion state of target to be tracked includes three kinds: unobstructed state, partial occlusion state and whole block shape
State;If 13 characteristic points of whole body all can be detected, it is determined that target to be tracked is in unobstructed state.If only
There is Partial Feature point to be detected, it is determined that target to be tracked is in partial occlusion state.If none characteristic point can
It is detected, it is determined that target to be tracked is in complete occlusion state.
4th step, according to different occlusion states, adjust cannot feature templates.
Here, if target to be tracked is in unobstructed state, continue using template A, in the target area, search not by
The most like part of the characteristic point that the characteristic point and template A at the position blocked include determines if similarity is greater than certain threshold value
Tracking is normal.If target to be tracked is in partial occlusion state, characteristic point is obtained, then determines the external of these characteristic points
Rectangle establishes template B (i.e. second feature template) using this rectangular area, then switches the feature templates of track algorithm,
It is switched to template B from original template A, continues the position for tracking and updating target to be tracked.When target occlusion to be tracked terminates
Afterwards, and from feature templates B template A is switched back into, to continue unobstructed tracking.It is hidden completely if target to be tracked is in
Gear state, then it is assumed that with losing, tracking process terminates target to be tracked.
The embodiment of the present application provides a kind of target tracker, which includes included each module and each module
Included each unit can be realized by the processor in computer equipment;Certainly it can also pass through specific logic circuit
It realizes;In the process of implementation, processor can be central processing unit (CPU), microprocessor (MPU), digital signal processor
(DSP) or field programmable gate array (FPGA) etc..
Fig. 4 is the composed structure schematic diagram of the embodiment of the present application target tracker, as shown in figure 4, described device 400 is wrapped
It includes: the first determining module 401, the second determining module 402, the first adjustment module 403 and the first search module 404, in which:
First determining module 401, the characteristic point at the position not being blocked for determining target to be tracked;
Second determining module 402, for determining the occlusion state of the target to be tracked according to the characteristic point;
The first adjustment module 403 adjusts the mesh to be tracked for the occlusion state according to the target to be tracked
Mark corresponding feature templates;
First search module 404, in the corresponding target area of the target to be tracked, search for and track with
The region that feature templates adjusted match;Wherein, the target area is for tracking the mesh target area to be tracked.
In above-mentioned apparatus, first determining module 401, comprising:
First determines submodule, for determining target to be tracked from image to be detected of input;
First detection sub-module obtains described not hidden for detecting the position of the target to be tracked not being blocked
The characteristic point at the position of gear.
In above-mentioned apparatus, second determining module 402, comprising:
Second determines submodule, if being more than or equal to first threshold for the characteristic point quantity, determines described to be tracked
The occlusion state of target is unobstructed state;
Third determines submodule, if being less than first threshold for the characteristic point quantity is more than or equal to second threshold, really
The occlusion state of the fixed target to be tracked is partial occlusion state;
4th determines submodule, if being less than second threshold for the characteristic point quantity, determines the target to be tracked
Occlusion state be complete occlusion state.
In above-mentioned apparatus, the first adjustment module 403, further includes: the first creation submodule, for according in nothing
The clarification of objective point to be tracked of occlusion state creates fisrt feature template;
Accordingly, first search module 404, comprising: the first search submodule, in the target pair to be tracked
In the target area answered, searches for and track the position that is not blocked and characteristic point that the fisrt feature template includes matches
First area.
In above-mentioned apparatus, the first adjustment module 403, comprising: the first adjustment submodule, if for it is described to
The occlusion state of track target is partial occlusion state, and fisrt feature template is adjusted to second feature template;Wherein, described second
The characteristic point that feature templates include is less than the characteristic point that the fisrt feature template includes;
Accordingly, first search module 404, comprising: the second search submodule, in the target pair to be tracked
In the target area answered, searches for and track the position that is not blocked and characteristic point that the second feature template includes matches
Second area.
In above-mentioned apparatus, the second search submodule, comprising:
First search unit, for searching for the position not being blocked in the corresponding target area of the target to be tracked
Characteristic point and the similarity of the second feature template characteristic point that includes be more than or equal to the second area of preset threshold.
In above-mentioned apparatus, described device further include:
Third determining module, if for the characteristic point that any region in the target area of the target to be tracked includes,
The characteristic point for including with the second feature template mismatches, and determines that the target to be tracked is lost, terminates tracking;Alternatively,
4th determining module determines the mesh to be tracked if being in complete occlusion state for the target to be tracked
Mark is lost, and tracking is terminated.
It should be noted that in the embodiment of the present application, if realized in the form of software function module above-mentioned target with
Track method, and when sold or used as an independent product, it also can store in a computer readable storage medium.Base
In such understanding, substantially the part that contributes to existing technology can be in other words for the technical solution of the embodiment of the present application
The form of software product embodies, which is stored in a storage medium, including some instructions to
So that one calculates equipment (can be personal computer or server etc.) and executes each embodiment the method for the application
It is all or part of.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (Read Only Memory,
ROM), the various media that can store program code such as magnetic or disk.In this way, the embodiment of the present application is not limited to any spy
Fixed hardware and software combines.
It should be noted that Fig. 5 is a kind of hardware entities schematic diagram of electronic equipment in the embodiment of the present application, such as Fig. 5 institute
To show, the hardware entities of the electronic equipment 500 include: processor 501, communication interface 502 and memory 503, wherein
The overall operation of the usual controlling electronic devices 500 of processor 501.
Communication interface 502 can make electronic equipment pass through network and other electronic equipments or server communication.
Memory 503 is configured to store the instruction and application that can be performed by processor 501, can also cache device to be processed
501 and electronic equipment 500 in each module it is to be processed or processed data (for example, image data, audio data, voice
Communication data and video communication data), flash memory (FLASH) or random access storage device (Random Access can be passed through
Memory, RAM) it realizes.
The description of medium stored above and apparatus embodiments, be with the description of above method embodiment it is similar, have same
The similar beneficial effect of embodiment of the method.For undisclosed technical detail in the application storage medium and apparatus embodiments, ask
Referring to the application embodiment of the method description and understand.
In several embodiments provided herein, it should be understood that disclosed device and method can pass through it
Its mode is realized.Apparatus embodiments described above are merely indicative, for example, the division of the unit, only
A kind of logical function partition, there may be another division manner in actual implementation, such as: multiple units or components can combine, or
It is desirably integrated into another system, or some features can be ignored or not executed.In addition, shown or discussed each composition portion
Mutual coupling or direct-coupling or communication connection is divided to can be through some interfaces, the INDIRECT COUPLING of equipment or unit
Or communication connection, it can be electrical, mechanical or other forms.
It is above-mentioned to can be or may not be and is physically separated as separation unit, the component as display can be,
Or it may not be physical unit;Both it can be located in one place, may be distributed over multiple network units;It can be according to reality
Border needs to select some or all of units to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the application can be fully integrated in one processing unit, it can also
To be each unit individually as a unit, can also be integrated in one unit with two or more units;It is above-mentioned
Integrated unit both can take the form of hardware realization, can also realize in the form of hardware adds SFU software functional unit.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can store in computer-readable storage medium, which exists
When execution, step including the steps of the foregoing method embodiments is executed;And storage medium above-mentioned includes: movable storage device, read-only deposits
Reservoir (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or
The various media that can store program code such as CD.
If alternatively, the above-mentioned integrated unit of the application is realized in the form of software function module and as independent product
When selling or using, it also can store in a computer readable storage medium.Based on this understanding, the application is implemented
Substantially the part that contributes to existing technology can be embodied in the form of software products the technical solution of example in other words,
The computer software product is stored in a storage medium, including some instructions are used so that computer equipment (can be with
Personal computer, server or network equipment etc.) execute each embodiment the method for the application all or part.
And storage medium above-mentioned includes: that movable storage device, ROM, RAM, magnetic or disk etc. are various can store program code
Medium.
The above, the only specific embodiment of the application, but the protection scope of the application is not limited thereto, it is any
Those familiar with the art within the technical scope of the present application, can easily think of the change or the replacement, and should all contain
Lid is within the scope of protection of this application.Therefore, the protection scope of the application should be based on the protection scope of the described claims.
Claims (10)
1. a kind of method for tracking target, which is characterized in that the described method includes:
Determine the characteristic point at the position of target to be tracked not being blocked;
According to the characteristic point, the occlusion state of the target to be tracked is determined;
According to the occlusion state of the target to be tracked, the corresponding feature templates of the target to be tracked are adjusted;
In the corresponding target area of the target to be tracked, searches for and track the area to match with feature templates adjusted
Domain;Wherein, the target area is for tracking the mesh target area to be tracked.
2. method as described in claim 1, which is characterized in that the position of the determination target to be tracked not being blocked
Characteristic point, comprising:
Target to be tracked is determined from image to be detected of input;
The position not being blocked for detecting the target to be tracked obtains the characteristic point at the position not being blocked.
3. method as described in claim 1, which is characterized in that it is described according to the characteristic point, determine the mesh to be tracked
Target occlusion state, comprising:
If the characteristic point quantity is more than or equal to first threshold, determine that the occlusion state of the target to be tracked is unobstructed shape
State;
If the characteristic point quantity is less than first threshold and is more than or equal to second threshold, determine the target to be tracked blocks shape
State is partial occlusion state;
If the characteristic point quantity is less than second threshold, determine that the occlusion state of the target to be tracked is to block shape completely
State.
4. method as described in claim 1, which is characterized in that the occlusion state according to the target to be tracked is adjusted
The corresponding feature templates of the whole target to be tracked, further includes: according to the clarification of objective point to be tracked for being in unobstructed state,
Create fisrt feature template;
Accordingly, described in the corresponding target area of the target to be tracked, it searches for and tracks and feature templates adjusted
The region to match, comprising: in the corresponding target area of the target to be tracked, search for and track the position that is not blocked with
The first area that the characteristic point that the fisrt feature template includes matches.
5. method as described in claim 4, which is characterized in that the occlusion state according to the target to be tracked is adjusted
The corresponding feature templates of the whole target to be tracked, comprising: if the occlusion state of the target to be tracked is partial occlusion shape
Fisrt feature template is adjusted to second feature template by state;Wherein, the characteristic point that the second feature template includes is less than described
The characteristic point that fisrt feature template includes;
Accordingly, the area that in the corresponding target area of the target to be tracked, search matches with the feature templates
Domain, comprising: in the corresponding target area of the target to be tracked, search for and track the position not being blocked and described second special
The second area that the characteristic point that sign template includes matches.
6. method as described in claim 5, which is characterized in that in the corresponding target area of the target to be tracked, search
The second area that the characteristic point that the position and the second feature template that Suo Wei is blocked include matches, comprising:
In the corresponding target area of the target to be tracked, the characteristic point and the second feature at the position not being blocked are searched for
The similarity for the characteristic point that template includes is more than or equal to the second area of preset threshold.
7. method as described in claim 5, which is characterized in that the method also includes:
If the characteristic point that any region includes in the target area of the target to be tracked, with the second feature template packet
The characteristic point contained mismatches, and determines that the target to be tracked is lost, terminates tracking;Alternatively,
If the target to be tracked is in complete occlusion state, determines that the target to be tracked is lost, terminate tracking.
8. a kind of target tracker, which is characterized in that described device includes: the first determining module, the second determining module, first
Adjust module and the first search module, in which:
First determining module, the characteristic point at the position not being blocked for determining target to be tracked;
Second determining module, for determining the occlusion state of the target to be tracked according to the characteristic point;
It is corresponding to adjust the target to be tracked for the occlusion state according to the target to be tracked for the first adjustment module
Feature templates;
First search module, in the corresponding target area of the target to be tracked, search for and track with after adjustment
The region that matches of feature templates;Wherein, the target area is for tracking the mesh target area to be tracked.
9. a kind of electronic equipment, including memory and processor, the memory are stored with the calculating that can be run on a processor
Machine program, which is characterized in that the processor realizes the target following of any one of claim 1 to 7 when executing described program
Step in method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program quilt
The step in any one of claim 1 to 7 method for tracking target is realized when processor executes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811355222.XA CN109712168A (en) | 2018-11-14 | 2018-11-14 | A kind of method for tracking target and electronic equipment, device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811355222.XA CN109712168A (en) | 2018-11-14 | 2018-11-14 | A kind of method for tracking target and electronic equipment, device and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109712168A true CN109712168A (en) | 2019-05-03 |
Family
ID=66254840
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811355222.XA Pending CN109712168A (en) | 2018-11-14 | 2018-11-14 | A kind of method for tracking target and electronic equipment, device and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109712168A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178201A (en) * | 2019-12-20 | 2020-05-19 | 南京理工大学 | Human body sectional type tracking method based on OpenPose posture detection |
CN111539283A (en) * | 2020-04-15 | 2020-08-14 | 上海摩象网络科技有限公司 | Face tracking method and face tracking equipment |
CN112184769A (en) * | 2020-09-27 | 2021-01-05 | 上海高德威智能交通系统有限公司 | Tracking abnormity identification method, device and equipment |
CN113139417A (en) * | 2020-11-24 | 2021-07-20 | 深圳云天励飞技术股份有限公司 | Action object tracking method and related equipment |
CN113139417B (en) * | 2020-11-24 | 2024-05-03 | 深圳云天励飞技术股份有限公司 | Action object tracking method and related equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103854293A (en) * | 2014-02-26 | 2014-06-11 | 奇瑞汽车股份有限公司 | Vehicle tracking method and device based on feature point matching |
CN105405151A (en) * | 2015-10-26 | 2016-03-16 | 西安电子科技大学 | Anti-occlusion target tracking method based on particle filtering and weighting Surf |
CN105654515A (en) * | 2016-01-11 | 2016-06-08 | 上海应用技术学院 | Target tracking method based on fragmentation and multiple cues adaptive fusion |
CN106548486A (en) * | 2016-11-01 | 2017-03-29 | 浙江大学 | A kind of unmanned vehicle location tracking method based on sparse visual signature map |
-
2018
- 2018-11-14 CN CN201811355222.XA patent/CN109712168A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103854293A (en) * | 2014-02-26 | 2014-06-11 | 奇瑞汽车股份有限公司 | Vehicle tracking method and device based on feature point matching |
CN105405151A (en) * | 2015-10-26 | 2016-03-16 | 西安电子科技大学 | Anti-occlusion target tracking method based on particle filtering and weighting Surf |
CN105654515A (en) * | 2016-01-11 | 2016-06-08 | 上海应用技术学院 | Target tracking method based on fragmentation and multiple cues adaptive fusion |
CN106548486A (en) * | 2016-11-01 | 2017-03-29 | 浙江大学 | A kind of unmanned vehicle location tracking method based on sparse visual signature map |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178201A (en) * | 2019-12-20 | 2020-05-19 | 南京理工大学 | Human body sectional type tracking method based on OpenPose posture detection |
CN111539283A (en) * | 2020-04-15 | 2020-08-14 | 上海摩象网络科技有限公司 | Face tracking method and face tracking equipment |
CN111539283B (en) * | 2020-04-15 | 2023-08-11 | 上海摩象网络科技有限公司 | Face tracking method and face tracking equipment |
CN112184769A (en) * | 2020-09-27 | 2021-01-05 | 上海高德威智能交通系统有限公司 | Tracking abnormity identification method, device and equipment |
CN112184769B (en) * | 2020-09-27 | 2023-05-02 | 上海高德威智能交通系统有限公司 | Method, device and equipment for identifying tracking abnormality |
CN113139417A (en) * | 2020-11-24 | 2021-07-20 | 深圳云天励飞技术股份有限公司 | Action object tracking method and related equipment |
CN113139417B (en) * | 2020-11-24 | 2024-05-03 | 深圳云天励飞技术股份有限公司 | Action object tracking method and related equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kristan et al. | The seventh visual object tracking vot2019 challenge results | |
CN109325456B (en) | Target identification method, target identification device, target identification equipment and storage medium | |
CN109712168A (en) | A kind of method for tracking target and electronic equipment, device and storage medium | |
Lee et al. | Human pose tracking in monocular sequence using multilevel structured models | |
Riaz et al. | Motion reconstruction using very few accelerometers and ground contacts | |
KR20160138062A (en) | Eye gaze tracking based upon adaptive homography mapping | |
Riazuelo et al. | Semantic visual slam in populated environments | |
CN109584302A (en) | Camera pose optimization method, device, electronic equipment and computer-readable medium | |
CN107944381B (en) | Face tracking method, face tracking device, terminal and storage medium | |
WO2021242445A1 (en) | Tracking multiple objects in a video stream using occlusion-aware single-object tracking | |
CN109948624A (en) | Method, apparatus, electronic equipment and the computer storage medium of feature extraction | |
Liu et al. | Robust multi-drone multi-target tracking to resolve target occlusion: A benchmark | |
CN117581272A (en) | Method and apparatus for team classification in sports analysis | |
He et al. | OVD-SLAM: An online visual SLAM for dynamic environments | |
JP2022521540A (en) | Methods and systems for object tracking using online learning | |
US8204274B2 (en) | Method and system for tracking positions of human extremities | |
Agarwal et al. | Boggart: Towards {General-Purpose} acceleration of retrospective video analytics | |
Park et al. | Extended keyframe detection with stable tracking for multiple 3d object tracking | |
CN101650776B (en) | Method and system for tracking position of human limbs | |
US10789778B1 (en) | Systems and methods for displaying augmented-reality objects | |
CN112417205A (en) | Target retrieval device and method and electronic equipment | |
WO2017139086A1 (en) | Performing multiple queries within a robust video search and retrieval mechanism | |
Kerdvibulvech et al. | Vision-based guitarist fingering tracking using a Bayesian classifier and particle filters | |
KR20140123399A (en) | Apparatus and Method of Body Parts Detection | |
Geng et al. | Reliable part guided multiple level attention learning for person re-identification |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |